paper-writing

Use this skill when working on:

• Research papers and journal articles
• Conference papers and proceedings
• Technical reports and white papers
• Thesis chapters and dissertations
• Literature reviews and survey papers
• Position papers and perspectives

For immediate guidance, see the task-specific workflows below. For comprehensive reference material, consult:

• references/REFERENCE.md - Complete writing guidelines and best practices
• references/STRUCTURE.md - Detailed templates for different paper types
• references/STYLE.md - Writing style and clarity guidelines
• assets/ - Ready-to-use paper templates

1. Planning Phase

Understand the Requirements

• Identify target venue (journal, conference, report)
• Check formatting requirements (APA, IEEE, ACM, Chicago, etc.)
• Note page limits, section requirements, and citation style
• Clarify submission deadlines and review process

Define the Research Question

• Formulate a clear, focused research question or hypothesis
• Ensure the question is specific, measurable, and answerable
• Identify the gap in existing knowledge or practice
• State the contribution your work will make

Create an Outline

• Draft section headings based on paper type (see Structure section)
• Allocate approximate space to each section
• Identify key points for each section
• Map evidence and references to sections

2. Writing Phase

Follow the Structured Approach

Work section-by-section, not necessarily in order. Many authors write in this sequence:

1. Methods (clearest, most concrete)
2. Results (present findings)
3. Introduction (frame the problem)
4. Discussion (interpret results)
5. Conclusion (summarize contributions)
6. Abstract (last, summarizes everything)

Section-Specific Guidance

Abstract (150-300 words)

• State the problem and motivation (1-2 sentences)
• Describe your approach/method (1-2 sentences)
• Summarize key results (2-3 sentences)
• State conclusions and implications (1-2 sentences)
• Make it self-contained (no citations, no undefined acronyms)

Introduction

• Hook: Why should anyone care? (1 paragraph)
• Context: What's the broader problem space? (1-2 paragraphs)
• Gap: What's missing in current solutions? (1 paragraph)
• Your contribution: What does this paper do? (1 paragraph)
• Paper organization: Brief roadmap (optional, 1 paragraph)

Related Work / Literature Review

• Group work thematically, not chronologically
• Compare and contrast approaches
• Identify limitations of existing work
• Position your work clearly vs. alternatives
• Be fair and accurate (don't strawman competitors)

Methodology / Approach

• Describe methods with enough detail for replication
• Justify design choices
• Define metrics and evaluation criteria
• Explain data collection and analysis procedures
• Include diagrams for complex processes

Results

• Present findings objectively without interpretation
• Use tables and figures effectively (see Visualization section)
• Report statistical significance where applicable
• Address both positive and negative results
• Organize by research question or hypothesis

Discussion

• Interpret results in context of research questions
• Compare with related work
• Explain unexpected findings
• Acknowledge limitations honestly
• Discuss implications for theory and practice

Conclusion

• Restate key contributions (1 paragraph)
• Summarize main findings (1 paragraph)
• Discuss broader implications (1 paragraph)
• Suggest future work (1 paragraph)
• End with a strong closing statement

3. Refinement Phase

First Revision: Structure and Argument

• Does each section serve its purpose?
• Is the argument logical and complete?
• Are transitions between sections smooth?
• Does evidence support all claims?
• Are counterarguments addressed?

Second Revision: Clarity and Style

• Remove jargon and define technical terms
• Eliminate redundancy and wordiness
• Use active voice for clarity (prefer "We analyzed" over "Analysis was performed")
• Ensure parallel structure in lists
• Check paragraph length (aim for 4-8 sentences)

Third Revision: Polish

• Check grammar, spelling, punctuation
• Verify all citations are formatted correctly
• Ensure figures/tables are referenced in text
• Number sections, equations, figures consistently
• Proofread carefully (reading aloud helps)

Tables

• Use for precise numerical comparisons
• Keep simple and readable (avoid excessive gridlines)
• Include clear column headers and units
• Caption goes above the table
• Reference in text before the table appears

Figures

• Use for trends, patterns, relationships
• Ensure axes are labeled with units
• Use readable fonts (at least 10pt in final size)
• Caption goes below the figure
• Make interpretable in grayscale (avoid color-only distinctions)

General Rules

• Every table/figure must be referenced in the text
• Each should be self-explanatory with its caption
• Number consecutively (Figure 1, Figure 2, etc.)
• Place close to first reference when possible

Structural Issues

• ❌ Burying the contribution in the middle
• ✅ State contribution clearly in introduction
• ❌ Results section that interprets rather than presents
• ✅ Keep results objective, interpret in discussion
• ❌ Conclusion that introduces new information
• ✅ Conclusion only synthesizes existing content

Writing Issues

• ❌ Passive constructions: "It was found that..."
• ✅ Active voice: "We found that..."
• ❌ Hedging excessively: "It seems to possibly suggest..."
• ✅ Be direct: "The results indicate..."
• ❌ Undefined acronyms and jargon
• ✅ Define terms on first use

Citation Issues

• ❌ Missing citations for claims
• ✅ Every factual claim needs a source
• ❌ Over-citing obvious facts
• ✅ Common knowledge doesn't need citation
• ❌ Citing without reading (citation padding)
• ✅ Cite only what you've actually read and verified

Computer Science / Engineering

• Emphasize reproducibility and implementation details
• Include complexity analysis where relevant
• Provide algorithm pseudocode or code snippets
• Compare against state-of-the-art baselines
• Make code/data available when possible

Natural Sciences

• Follow strict IMRAD structure (Intro, Methods, Results, Discussion)
• Report statistical power and effect sizes
• Include detailed experimental protocols
• Address confounding variables
• Report null results honestly

Social Sciences

• Contextualize within theoretical frameworks
• Justify sampling and participant selection
• Report demographic information
• Address potential biases
• Discuss generalizability limitations

Humanities

• Develop clear thesis statement
• Support arguments with textual evidence
• Engage with scholarly debates
• Use close reading and analysis
• Contextualize within historical/cultural frameworks

When helping users write papers, prioritize natural conversation over rapid-fire questions:

1. Start simple: Ask ONE opening question to understand their situation

- "What's your paper about?" (topic/working title)

- OR "What stage are you at?" (if they're already focused)

1. Listen and follow up: Based on their answer, ask the next most relevant question

- If they're starting: Ask about target venue or research gap

- If they're revising: Ask which section needs work

- If they're stuck: Ask what specific challenge they're facing

1. Build context progressively: Let information emerge naturally through dialogue

1. Provide targeted help: Focus on their immediate need, not all possible topics

1. Maintain author voice: Edit to improve, not to rewrite in a different style

1. Explain recommendations: Help them understand why changes strengthen their paper

1. Load references as needed: Use reference documents for deeper guidance only when relevant

This skill includes detailed reference materials for deeper guidance:

• Load references/REFERENCE.md for comprehensive writing guidelines, style guides, and detailed best practices
• Load references/STRUCTURE.md for complete templates for different paper types (journal article, conference paper, thesis, etc.)
• Load references/STYLE.md for in-depth guidance on academic writing style, word choice, and clarity
• Check assets/ for ready-to-use templates in various formats

Only load these materials when needed to conserve context. For most tasks, the guidance above is sufficient.

Basic use: Follow the Core Workflow for writing or revising papers

Intermediate use: Consult reference materials for specific sections or challenges

Advanced use: Use templates from assets/ and customize for specific venues

Before considering a paper complete, verify:

• [ ] Clear research question or contribution stated in introduction
• [ ] All claims supported by evidence or citations
• [ ] Logical flow from section to section
• [ ] Methods sufficiently detailed for replication
• [ ] Results presented objectively
• [ ] Discussion interprets results and addresses limitations
• [ ] All figures/tables referenced in text and have clear captions
• [ ] Citations formatted consistently
• [ ] Abstract accurately summarizes the paper
• [ ] No typos or grammatical errors
• [ ] Meets venue requirements (format, length, style)

Example 1: Progressive discovery—starting fresh

```

User: Help me write a research paper.

Claude: What's your paper about? (topic, working title, or research question)

User: It's about federated learning for edge devices.

Claude: Got it—that's a focused topic. Are you at the stage of having completed

research and needing to write it up, or are you still planning the work?

User: I've done the research and have preliminary results. I just don't know

where to start.

Claude: Perfect—let me help you structure this. What's your target venue?

(e.g., a specific conference, journal, or is this a thesis chapter?)

```

Notice: One question asked, then the next, building naturally from their answers.

---

Example 2: Targeted revision (existing draft)

```

User: My introduction feels weak and I'm not sure why.

Claude: I'd like to take a look. Can you share what you have, or tell me

what the main challenge is? (e.g., doesn't grab attention, doesn't explain

why this work matters, etc.)

User: I think I'm not making the research gap clear enough. Here's what I wrote...

Claude: [Reads their draft]

Looking at your introduction, I see you've identified the problem nicely,

but you're right—the gap (what's missing in current solutions) isn't

explicit. Here's what I'd suggest...

```

Notice: Start with curiosity about their specific challenge, then provide

targeted feedback on the actual content.

---

Example 3: Stuck in the middle

```

User: I have my methods section written, but I'm stuck on results.

Claude: What's tripping you up about the results section specifically?

(e.g., how to organize findings, whether to include raw data, interpretation

vs. presentation, etc.)

User: I have a lot of data. I don't know what to include and how to present it

so it's not overwhelming.

Claude: That's a common challenge. Let me help you think through data selection

and visualization...

```

Notice: Rather than lecturing about results sections in general, we drill down

into their specific problem first.

```

Writing Process

• Write regularly in focused sessions (2-3 hours)
• Don't edit while drafting (separate creation from refinement)
• Get feedback early and often
• Read papers in your target venue for style/structure

Revision Strategy

• Take breaks between revision passes
• Read paper aloud to catch awkward phrasing
• Get fresh eyes (colleagues, advisors)
• Use revision checklist systematically

Time Management

• Allocate 40% planning, 30% writing, 30% revision
• Build in buffer time before deadlines
• Start with rough drafts, refine iteratively
• Don't aim for perfection in first draft

Remember: All good papers go through multiple revisions. Writing is rewriting.